OpenAI has disclosed that more than one million people each week show signs of suicidal intent when using its ChatGPT platform. According to the company’s latest blog post, these conversations often contain “explicit indicators of potential suicidal planning or intent,” underscoring the growing link between artificial intelligence and global mental health challenges.
Growing Need for Online Mental Health Support
The report highlights how online mental health support has become increasingly critical as AI-driven chat services are used by millions worldwide. OpenAI revealed that about 0.07% of users in a given week—roughly 560,000 people—show potential signs of mental health emergencies like psychosis or mania. While these figures are preliminary, they point to a concerning trend: people are turning to chatbots in moments of emotional crisis.
The announcement comes amid rising scrutiny of how AI systems handle mental health issues. A recent lawsuit by the family of a teenager who died by suicide after frequent interactions with ChatGPT has placed OpenAI under public and legal pressure. Additionally, the U.S. Federal Trade Commission (FTC) has launched an investigation into AI developers’ practices around safeguarding young users.
GPT-5 Brings New Safety Measures and Suicide Prevention Resources
OpenAI said its latest GPT-5 model shows a marked improvement in safety and reliability during sensitive conversations. In internal evaluations involving over 1,000 self-harm and suicide-related chats, GPT-5 achieved a 91% compliance rate with desired behaviors—up from 77% in its previous version.
As part of its safety overhaul, OpenAI integrated suicide prevention resources directly into the chatbot, including expanded access to crisis hotlines and reminders encouraging users to take breaks. The company collaborated with 170 medical professionals from its Global Physician Network to review 1,800 mental health scenarios and refine how GPT-5 responds to users in distress.
Expert Concerns About AI and Mental Health
Mental health experts continue to warn about overreliance on chatbots for emotional support. While mental health chat services can offer immediate help, professionals emphasize that they should not replace therapy or crisis counseling. AI tools risk reinforcing users’ harmful thoughts—a behavior known as “sycophancy”—if not properly trained and monitored.
OpenAI’s CEO, Sam Altman, acknowledged the company’s cautious approach, stating that earlier restrictions were meant to minimize harm. He announced plans to relax certain limitations now that safety tools have improved, signaling a new phase for AI’s role in wellness and emotional care.
The Bigger Picture
“Mental health symptoms and emotional distress are universal,” OpenAI stated in its blog post. “As our user base grows, it’s expected that some conversations will involve these sensitive topics.”
For individuals in crisis, human support remains vital. In the U.S., call or text 988 to reach the National Suicide Prevention Lifeline, visit 988lifeline.org, or text HOME to 741741 for help. In the U.K. and Ireland, contact Samaritans at 116 123 or email jo@samaritans.org. In Australia, call Lifeline at 13 11 14. For more international helplines, visit befrienders.org.
If you or someone you know is struggling, please seek professional support or use available crisis helplines. For travel and relaxation ideas to reduce stress, explore our travel booking site.
This story was first reported by The Guardian. Read the full article here.
